Skip to content
DenseChip
Learn
PracticeCode labInterview

DenseChip

Navigate

Menu

LearnPracticeCode labInterview

Tools

All toolsSV editorCode repoResume
DenseChip

Practice, courses, and chip-design tools in one focused workspace.

LearnPracticeCode labInterviewToolsLearning paths

© 2026 DenseChip

AboutContactPrivacyTerms
Skip to lesson

Course contents

  • How to Use This Course
  • Protocol Metrics Reference
  • Protocols Glossary
  • Reading Path
  • Protocol Fundamentals61
  • AMBA Family (APB / AHB / AXI)61
  • Coherence Fabrics (ACE / CHI)61
    • ACE Coherent Transactions
    • Snoop & Cache Maintenance
    • CHI Topology Basics
    • Coherence Fabric Debug
    • Coherence Fabrics (ACE / CHI): Tricky Q&A
  • Memory Interfaces (DDR / LPDDR / HBM)61
  • PCIe & CXL61
  • Embedded Peripherals (I2C / SPI / UART)61
  • High-Speed I/O (USB / Ethernet / MIPI)61
  • Protocol Verification & Compliance61
  • Protocols Interview Prep5
  • Interface Protocols Cheatsheet
Loading lesson

‹Coherence Fabrics (ACE / CHI)
  1. Topics
  2. Bus Protocols
  3. Coherence Fabrics (ACE / CHI)
  4. Coherence Fabrics (ACE / CHI): Tricky Q&A

Interface Protocols · All levels

Coherence Fabrics (ACE / CHI): Tricky Q&A

Senior interview and review questions for Coherence Fabrics (ACE / CHI).

On this page

    Section Q&A bank

    Use these drills after completing all topics in Coherence Fabrics (ACE / CHI). Answer with layer, transaction, metric, artifact, and regression.

    What is false sharing?

    Diagram
    [INT][PROT][COHERENCE-FABRICS]
    
    Q: What is false sharing?
    
    A:
    Different variables on the same cache line cause coherence traffic without logical data sharing.
    
    FOLLOW-UP TRAP: Fixing with a faster NoC only.

    What does the Shared state mean in MESI?

    Diagram
    [INT][PROT][COHERENCE-FABRICS]
    
    Q: What does the Shared state mean in MESI?
    
    A:
    Line is valid and may be shared; reads allowed, writes require upgrade/invalidate sequence.
    
    FOLLOW-UP TRAP: Confusing Shared with Exclusive.

    Why does CHI separate request, response, data, and snoop flows?

    Diagram
    [INT][PROT][COHERENCE-FABRICS]
    
    Q: Why does CHI separate request, response, data, and snoop flows?
    
    A:
    To scale many-core systems with independent channels and avoid protocol deadlock.
    
    FOLLOW-UP TRAP: Treating CHI as 'AXI with snoops'.

    Q&A drill guide

    Diagram
    LAYER -> TRANSACTION -> METRIC -> ARTIFACT -> FIX -> REGRESSION

    Sketch while answering

    Diagram
    COHERENCE TRAFFIC FLOW
    
    RN issues coherent read
       -> HN looks up directory
       -> snoops to sharers
       -> data + state update returned
    
    False sharing: different variables, same cache line -> coherence storm.

    Common traps

    • If snoop latency high, check home node placement and directory policy.

    • If ordering bug, run litmus sequences before microarch changes.

    • If traffic storm, profile cache line sharing in software layout.

    Key takeaways

    • Connect every protocol claim to a transaction identity and measurable metric.

    • Store the artifact (waveform, log, counter) next to every signoff decision.

    Common pitfalls

    • Debugging timeouts without finding the first bad transaction.

    • Quoting peak bus width without payload efficiency and retry overhead.

    • Treating VIP compliance as a substitute for system integration replay.

    PreviousNext: Memory Interfaces (DDR / LPDDR / HBM)

    Up next

    Up nextMemory Interfaces (DDR / LPDDR / HBM)

    Sign in to save progress

    Practice this lessonReport an error

    See also

    • ACE Coherent Transactions
    • Snoop & Cache Maintenance
    • CHI Topology Basics